Challenges, opportunities and recent advances in near infrared spectroscopy applications for monitoring blend uniformity in the continuous manufacturing of solid oral dosage forms

Application of adaptive chaotic dung beetle optimization algorithm to near-infrared spectral model transfer

A new transfer approach was proposed to share calibration models of the hexamethylenetetramine-acetic acid solution for studying hexamethylenetetramine concentration values across different near-infrared (NIR) spectrometers. This approach combines Savitzky-Golay first derivative (S_G_1) and orthogonal signal correction (OSC) preprocessing, along with feature variable optimization using an adaptive chaotic dung beetle optimization (ACDBO) algorithm. The ACDBO algorithm employs tent chaotic mapping and a nonlinear decreasing strategy, enhancing the balance between global and local search capabilities and increasing population diversity to address limitations observed in traditional dung beetle optimization (DBO). Validated using the CEC-2017 benchmark functions, the ACDBO algorithm demonstrated superior convergence speed, accuracy, and stability. In the context of a partial least squares (PLS) regression model for transferring hexamethylenetetramine-acetic acid solutions using NIR spectroscopy, the ACDBO algorithm excelled over alternative methods such as uninformative variable elimination, competitive adaptive reweighted sampling, cuckoo search, grey wolf optimizer, differential evolution, and DBO in efficiency, accuracy of feature variable selection, and enhancement of model predictive performance. The algorithm attained outstanding metrics, including a determination coefficient for the calibration set (Rc2) of 0.99999, a root mean square error for the calibration set (RMSEC) of 0.00195%, a determination coefficient for the validation set (Rv2) of 0.99643, a root mean squared error for the validation set (RMSEV) of 0.03818%, residual predictive deviation (RPD) of 16.72574. Compared to existing OSC, slope and bias correction (S/B), direct standardization (DS), and piecewise direct standardization (PDS) model transfer methods, the novel strategy enhances the accuracy and robustness of model predictions. It eliminates irrelevant background information about the hexamethylenetetramine concentration, thereby minimizing the spectral discrepancies across different instruments. As a result, this approach yields a determination coefficient for the prediction set (Rp2) of 0.96228, a root mean squared error for the prediction set (RMSEP) of 0.12462%, and a relative error rate (RER) of 17.62331, respectively. These figures closely follow those obtained using DS and PDS, which recorded Rp2, RMSEP, and RER values of 0.97505, 0.10135%, 21.67030, and 0.98311, 0.08339%, 26.33552, respectively. Unlike conventional methods such as OSC, S/B, DS, and PDS, this novel approach does not require the analysis of identical samples across different instruments. This characteristic significantly broadens its applicability for model transfer, which is particularly beneficial for transferring specific measurement samples.

THz-TDS as a PAT tool for monitoring blend homogeneity in pharmaceutical manufacturing of solid oral dosage forms: A proof-of-concept study

The process analytical technology (PAT) framework is well established and integral to facilitate process understanding, enable a transition from batch to continuous manufacturing, and improve product quality. Near-infrared (NIR) spectroscopy has been established as a standard PAT tool for many process analytical challenges, including monitoring powder blend homogeneity. However, alternative technologies for monitoring powder blending are of interest due to the importance of the blending step in manufacturing solid oral dosage forms. Terahertz time-domain spectroscopy (THz-TDS) is therefore explored in this study as an alternative tool for monitoring blend homogeneity with the potential for endpoint control in a batch blending process. Powder blends of microcrystalline cellulose (MCC) and dibasic calcium phosphate dihydrate and blends of MCC and granulated α-lactose monohydrate were investigated non-invasively at various compositions using THz-TDS in transmission mode for acquiring spectra from samples enclosed in the blending container. It was found that attenuation- and phase-related parameters acquired with THz-TDS could reliably resolve physical changes related to the homogeneity of the blend. Further evaluations revealed that changes in the bulk density of the blend, in addition to the intrinsic optical properties of the materials, played a critical role in the observed trends for both systems. In contrast, the scattering contribution of the powder was mainly crucial for the attenuation-related parameter in blends with materials of high refractive indices. Finally, THz-TDS measurements were acquired throughout a blending process mimicking a continuous acquisition. The method could follow blending dynamics and resulted in reasonable predictive errors of the content of 0.5 - 2.5 %. Relative standard deviations for high content blends (20 %) were acceptable (3 - 7 %) whereas at low contents (5 %) significantly higher values (9 - 35 %) were found. Based on these findings, THz-TDS is a feasible PAT tool for monitoring blend homogeneity and controlling high content blend processes, although precision and accuracy is considered to improve with a more suitable interface.

Implementing multiblock techniques in a full-scale plant scenario: On-line prediction of quality parameters in a continuous process for different acrylonitrile butadiene styrene (ABS) products

BACKGROUND

The study explores the challenges of handling multiblock data of different natures (process and NIR sensors) for on-line quality prediction in a full-scale plant scenario, namely a plant operating in continuous on an industrial scale and producing different grade Acrylonitrile Butadiene Styrene (ABS) products. This environment is an ideal scenario to evaluate the use of multiblock data analysis methods, which can enhance data interpretation, visualization, and predictive performances. In particular, a novel multiblock extension of Locally Weighted PLS has been proposed by the authors, namely Locally Weighted Multiblock Partial Least Squares (LW-MB-PLS). Response-Oriented Sequential Alternation (ROSA) has also been employed to evaluate the diverse block relevance for the prediction of two quality parameters associated with the polymer. Data are split in blocks both according to sensor type and different plant sections, and different models have been built by incremental addition of data blocks to evaluate if early estimation of product quality is feasible.

RESULTS

ROSA method showed promising predictive performance for both quality parameters, highlighting the most influential plant sections through the selection of data blocks. The results suggested that both early and late-stage sensors play crucial roles in predicting product quality. A reasonable estimation of quality parameters before production completion has been achieved. On the other hand, the proposed LW-MB-PLS, while comparable in predictive performances, allowed reducing systematic prediction errors for specific products.

SIGNIFICANCE

This study contributes valuable insights for continuous production processes, aiding plant operators and paving the way for advancements in online quality prediction and control. Furthermore, it is implemented as a locally weighted extension of MB-PLS.

Real-time release testing of in vitro dissolution and blend uniformity in a continuous powder blending process by NIR spectroscopy and machine vision

Continuous manufacturing is gaining increasing interest in the pharmaceutical industry, also requiring real-time and non-destructive quality monitoring. Multiple studies have already addressed the possibility of surrogate in vitro dissolution testing, but the utilization has rarely been demonstrated in real-time. Therefore, in this work, the in-line applicability of an artificial intelligence-based dissolution surrogate model is developed the first time. NIR spectroscopy-based partial least squares regression and artificial neural networks were developed and tested in-line and at-line to assess the blend uniformity and dissolution of encapsulated acetylsalicylic acid (ASA) - microcrystalline cellulose (MCC) powder blends in a continuous blending process. The studied blend is related to a previously published end-to-end manufacturing line, where the varying size of the ASA crystals obtained from a continuous crystallization significantly affected the dissolution of the final product. The in-line monitoring was suitable for detecting the variations in the ASA content and dissolution caused by the feeding of ASA with different particle sizes, and the at-line predictions agreed well with the measured validation dissolution curves (f2 = 80.5). The results were further validated using machine vision-based particle size analysis. Consequently, this work could contribute to the advancement of RTRT in continuous end-to-end processes.

Continuous manufacturing of pharmaceutical products: A density-insensitive near infrared method for the in-line monitoring of continuous powder streams

Near infrared (NIR) spectroscopy is a valuable analytical technique for monitoring chemical composition of powder blends in continuous pharmaceutical processes. However, the variation in density captured by NIR during spectral collection of dynamic powder streams at different flow rates often reduces the performance and robustness of NIR models. To overcome this challenge, quantitative NIR measurements are commonly collected across all potential manufacturing conditions, including multiple flow rates to account for the physical variations. The utility of this approach is limited by the considerable quantity of resources required to run and analyze an extensive calibration design at variable flow rates in a continuous manufacturing (CM) process. It is hypothesized that the primary variation introduced to NIR spectra from changing flow rates is a change in the density of the powder from which NIR spectra are collected. In this work, powder stream density was used as an efficient surrogate for flow rate in developing a quantitative NIR method with enhanced robustness against process rate variation. A density design space of two process parameters was generated to determine the conditions required to encompass the apparent density and spectral variance from increases in process rate. This apparent density variance was included in calibration at a constant low flow rate to enable the development of a density-insensitive NIR quantitative model with limited consumption of materials. The density-insensitive NIR model demonstrated comparable prediction performance and flow rate robustness to a traditional NIR model including flow rate variation ("gold standard" model) when applied to monitoring drug content in continuous runs at varying flow rates. The proposed platform for the development of in-line density-insensitive NIR methods is expected to facilitate robust analytical model performance across variable continuous manufacturing production scales while improving the material efficiency over traditional robust modeling approaches for calibration development.

The Use of a Closed Feed Frame for the Development of Near-Infrared Spectroscopic Calibration Model to Determine Drug Concentration

PURPOSE

This study evaluates the use of the closed feed frame as a material sparing approach to develop near-infrared (NIR) spectroscopic calibration models for monitoring blend uniformity. The effect of shear induced by recirculation on NIR spectra was also studied.

METHODS

Calibration models were developed using NIR spectra obtained in the closed feed frame for two cases. For case 2, blends that flowed through the open feed frame were predicted with the model. The shear effect of the feed frame on the blends was assessed through the characterization of powder properties before and after recirculation.

RESULTS

The physical characterization of the blends confirmed that the powder properties were not altered after recirculation within the closed feed frame. Both calibration models provided highly accurate predictions of the test sets with low bias (0.03% w/w and -0.06% w/w) and relative standard error of prediction (1.9% and 3.7%), respectively. The predictive performance of the calibration models was not affected by the shear effect.

CONCLUSION

Recirculation within the closed feed frame did not change the physical properties of the blends studied. The prediction of blends flowing through the open feed frame was possible with a calibration model developed in the closed feed frame. The closed feed frame could reduce the materials needed to develop calibration models by more than 90%.

Influence of powder stream density on near infrared measurements upon scale-up of a simulated continuous process

Near-infrared (NIR) spectroscopy is a powerful process analytical tool for monitoring chemical constituents in continuous pharmaceutical processes. However, the density variation introduced when quantitative NIR measurements are performed on powder streams at different flow rates is a potential source of a lack of model robustness. Since different flow rates are often required to meet the production requirements (e.g., during scale-up) of a continuous process, the development of efficient strategies to characterize, understand, and mitigate the impact of powder density on NIR measurements is highly desirable. This study focused on assessing the effect of powder physical variation on NIR by enabling the in-line characterization of powder stream density in a simulated continuous system. The in-line measurements of powder stream density were facilitated through a unique analytical interface to a flowing process. Powder streams delivered at various design levels of flow rate and tube angle were monitored simultaneously by NIR diffuse reflectance spectroscopy, live imaging, and dynamic mass characterization. Statistical analysis and multivariate modeling confirmed powder density as a significant source of spectral variability due to flow rate. Besides providing broader process understanding, results elucidated potential mitigation strategies to facilitate effective continuous process scale-up while ensuring NIR model robustness against density.

Application of a wavelength angle mapper for variable selection in iterative optimization technology predictions of drug content in pharmaceutical powder mixtures

Process analytical technology (PAT) is an essential tool within pharmaceutical manufacturing to ensure consistent quality and maintain process control. Near-infrared (NIR) spectroscopy is one of the most popular PAT techniques, particularly for monitoring active pharmaceutical ingredient (API) concentrations. To interpret the spectral outputs of NIR spectroscopy, advanced multivariate models are required. Calibration-free models such as iterative optimization technology (IOT) algorithms are increasingly of interest, due primarily to their reduced material and time burdens. Variable/wavelength selection is a common method to improve prediction performance and robustness for IOT by focusing on spectral regions with the most relevant information. However, currently proposed wavelength selection approaches rely on training sets for optimization, therefore reducing or removing the advantages of IOT over empirical calibration-dependent models. In this work, a true calibration-free wavelength selection method is proposed based on measuring the difference between individual wavelengths of a mixture spectra and the net analyte signals via a wavelength angle mapper (WAM). An extension of the WAM utilizing a spectral window of wavelength instead of individual wavelengths, called SWAM, was also developed. However, the SWAM method does require a small training set to optimize wavelength selection parameters. The WAM and SWAM methods showed similar prediction performance for API in pharmaceutical powder blends when compared against other calibration-dependent models and the base IOT algorithm.

Advancing non-destructive analysis of 3D printed medicines

Pharmaceutical 3D printing (3DP) has attracted significant interest over the past decade for its ability to produce personalised medicines on demand. However, current quality control (QC) requirements for traditional large-scale pharmaceutical manufacturing are irreconcilable with the production offered by 3DP. The US Food and Drug Administration (FDA) and the UK Medicines and Healthcare Products Regulatory Agency (MHRA) have recently published documents supporting the implementation of 3DP for point-of-care (PoC) manufacturing along with regulatory hurdles. The importance of process analytical technology (PAT) and non-destructive analytical tools in translating pharmaceutical 3DP has experienced a surge in recognition. This review seeks to highlight the most recent research on non-destructive pharmaceutical 3DP analysis, while also proposing plausible QC systems that complement the pharmaceutical 3DP workflow. In closing, outstanding challenges in integrating these analytical tools into pharmaceutical 3DP workflows are discussed.

Veering to a Continuous Platform of Fused Deposition Modeling Based 3D Printing for Pharmaceutical Dosage Forms: Understanding the Effect of Layer Orientation on Formulation Performance

Three-dimensional (3D) printing of pharmaceuticals has been centered around the idea of personalized patient-based 'on-demand' medication. Fused deposition modeling (FDM)-based 3D printing processes provide the capability to create complex geometrical dosage forms. However, the current FDM-based processes are associated with printing lag time and manual interventions. The current study tried to resolve this issue by utilizing the dynamic z-axis to continuously print drug-loaded printlets. Fenofibrate (FNB) was formulated with hydroxypropyl methylcellulose (HPMC AS LG) into an amorphous solid dispersion using the hot-melt extrusion (HME) process. Thermal and solid-state analyses were used to confirm the amorphous state of the drug in both polymeric filaments and printlets. Printlets with a 25, 50, and 75% infill density were printed using the two printing systems, i.e., continuous, and conventional batch FDM printing methods. Differences between the two methods were observed in the breaking force required to break the printlets, and these differences reduced as the infill density went up. The effect on in vitro release was significant at lower infill densities but reduced at higher infill densities. The results obtained from this study can be used to understand the formulation and process control strategies when switching from conventional FDM to the continuous printing of 3D-printed dosage forms.

Near infrared spectroscopy quantification based on Bi-LSTM and transfer learning for new scenarios

This study proposed a deep transfer learning methodology based on an improved Bi-directional Long Short-Term Memory (Bi-LSTM) network for the first time to address the near infrared spectroscopy (NIR) model transfer issue between samples. We tested its effectiveness on two datasets of manure and polyglutamic acid (γ-PGA) solution, respectively. First, the optimal primary Bi-LSTM networks for cattle manure and the first batch of γ-PGA were developed by ablation experiments and both proved to outperform one-dimensional convolutional neural network (1D-CNN), Partial Least Square (PLS) and Extreme Learning Machine (ELM) models. Then, two types of transfer learning approaches were carried out to determine model transferability to non-homologous samples. For poultry manure and the second batch of γ-PGA, the obtained predicting results verified that the second approach of fine-tuning Bi-LSTM layers and re-training FC layers transcended the first approach of fixing Bi-LSTM layers and only re-training FC layers by reducing the RMSEP_target of 23.4275% and 50.7343%, respectively. Finally, comparisons with fine-tuning 1D-CNN and other traditional model transfer methods further identified the superiority of the proposed methodology with exceeding accuracy and smaller variation, which decreased RMSEP_target of poultry manure and the second batch of γ-PGA of 7.2832% and 48.1256%, 67.1117% and 80.6924% when compared to that acquired by fine-tuning 1D-CNN, Tradaboost-ELM and CCA-PLS which were the best of five traditional methods, respectively. The study demonstrates the potential of the Fine-tuning-Bi-LSTM enabled NIR technology to be used as a simple, cost effective and reliable detection tool for a wide range of applications under various new scenarios.

NIR Spectroscopy as an Online PAT Tool for a Narrow Therapeutic Index Drug: Toward a Platform Approach Across Lab and Pilot Scales for Development of a Powder Blending Monitoring Method and Endpoint Determination

An online near-infrared (NIR) spectroscopy platform system for real-time powder blending monitoring and blend endpoint determination was tested for a phenytoin sodium formulation. The study utilized robust experimental design and multiple sensors to investigate multivariate data acquisition, model development, and model scale-up from lab to manufacturing. The impact of the selection of various blend endpoint algorithms on predicted blend endpoint (i.e., mixing time) was explored. Spectral data collected at two process scales using two NIR spectrometers was incorporated in a single (global) calibration model. Unique endpoints were obtained with different algorithms based on standard deviation, average, and distributions of concentration prediction for major components of the formulation. Control over phenytoin sodium's distribution was considered critical due to its narrow therapeutic index nature. It was found that algorithms sensitive to deviation from target concentration offered the simplest interpretation and consistent trends. In contrast, algorithms sensitive to global homogeneity of active and excipients yielded the longest mixing time to achieve blending endpoint. However, they were potentially more sensitive to subtle uniformity variations. Qualitative algorithms using principal component analysis (PCA) of spectral data yielded the prediction of shortest mixing time for blending endpoint. The hybrid approach of combining NIR data from different scales presents several advantages. It enables simplifying the chemometrics model building process and reduces the cost of model building compared to the approach of using data solely from commercial scale. Success of such a hybrid approach depends on the spectroscopic variability captured at different scales and their relative contributions in the final NIR model.